Lithography Inspection Condition Prediction via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining inspection conditions for patterns on substrates are ineffective when the correspondence between measured values and inspection conditions is unclear, limiting the ability to accurately inspect patterns formed using lithography techniques.
Innovation Solution
An information processing apparatus that uses machine learning to acquire inspection conditions by inputting state information of the lithography apparatus, leveraging learning data that includes information about the state of the apparatus and previously applied inspection conditions to predict optimal inspection parameters for new patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inspection is performed on all shot areas on the substrate, then inspection accuracy and quality are improved, but inspection time increases significantly
Solution Approach 1:
The patent applies preliminary action by determining inspection conditions before the actual inspection based on measured values obtained during the lithography process. The system predicts which shot areas require inspection and under what conditions, allowing the inspection to be performed efficiently without needing to inspect all areas manually, thus reducing inspection time while maintaining accuracy.
Solution Approach 2:
The patent implements feedback by using measured values from the lithography process (such as exposure dose, focus, and alignment information) to determine inspection conditions. This feedback loop allows the system to automatically adjust inspection parameters based on actual process data, enabling selective inspection of critical areas only, thereby reducing overall inspection time while maintaining high measurement precision.
2Productivity
If inspection conditions are determined based on measured values from the lithography process, then inspection efficiency is improved, but the method fails when correspondence between measured values and inspection conditions is unclear
Solution Approach 1:
The patent applies parameter changes by adapting the determination of inspection conditions to the specific characteristics of each shot area. Instead of using fixed inspection conditions, the system adjusts parameters such as inspection location, timing, and criteria based on the measured values from the lithography process for each individual shot area, ensuring reliable inspection condition determination even when general correspondence rules are unclear.
3Loss of time
If selective inspection is performed on only some shot areas, then inspection time is reduced, but inspection coverage and quality control are compromised
Solution Approach 1:
The patent applies local quality by determining that different shot areas require different inspection conditions based on their specific characteristics. Instead of uniform inspection, the system identifies which shot areas have critical parameters (such as those with abnormal exposure or alignment) and applies enhanced inspection conditions to those specific locations, ensuring quality control is maintained where needed while reducing overall inspection time.
Data Source
AI summary
An information processing apparatus for acquiring an inspection condition for performing an inspection on a pattern formed by a lithography apparatus that forms a pattern on a substrate with an original includes an acquisition unit configured to acquire a second inspection condition to be applied in a case where an inspection is performed on a second pattern by inputting third information indicating a state of the lithography apparatus acquired when the second pattern is formed to a model, wherein the model is acquired by machine learning with learning data including first information indicating a state of the lithography apparatus acquired when a first pattern is formed and second information indicating a first inspection condition applied when an inspection is performed on the first pattern.